Research On Loop Closure And Graph Optimization Based On Key Frames

With the rapid development of computer technology,intelligent robots play an increasingly important role in human life,and related research has become a hot research direction of scholars.To reflect the intelligence of a robot,it needs to be able to walk freely in any unknown environment like a human being.To achieve this function,SLAM is one of the key technologies.The full name of SLAM(Simultaneously Localization and Mapping)is instant positioning and composition.The sensors are used to perceive the environmental information,and the SLAM algorithm is used to construct map and mobile robot location in an unknown environment.As an algorithm that has developed for many years,the focused areas of SLAM have evolved from the early use of ranging sensors to acquire 2D environmental information and incremental maps based on probabilistic model calculations to the use of cameras to capture more information in 3D environments and maps construction based on optimized graph structures.All these are indicating that the field has not yet matured and there are still many areas for improvement and research.In this paper,the improved algorithm improve the problems of map optimization and loop closure methods in 2D Graph-based SLAM,and combine experiments to prove the significance.The specific work is as follows:(1)Analyze the importance and application status of SLAM algorithm in intelligent robots,introduce the development of SLAM algorithm,the present situation at domestic and abroad,and relevant basic knowledge of SLAM algorithms.In-depth study and analysis of graph-based SLAM algorithm,and use GraphSLAM as an example to illustrate the theory and solution of SLAM algorithm.(2)Introducing the current general graph optimization framework,inferring the process of solving the optimal solution,aiming at solving the problem of complicated computation due to data accumulation in the information matrix of the solution formula,constructing key frames and reducing information matrix element calculations.Secondly,based on keyframes,a global graph is construct to improve the framework of the graph optimization algorithm in the SLAM system.A global graph-oriented graph optimization framework is proposed with SLAM system to improve the effectiveness and speed of the optimization algorithm.Finally,combined with simulation experiments to illustrate the feasibility of the improved algorithm.(3)Based on CSM(Correspondence Scan Matching)algorithm for loop closure,aiming at insufficient matching and misdetection of the original algorithm,an method is proposed to improve the matching accuracy with key frames.Because the improved method of the key frame will reduce the calculation speed of the algorithm,on the basis of the improvement method,it proposes to improve the filter method of the candidate set of loop closure to improve the detection speed of the algorithm for ensuring that the accuracy rate does not decrease.Simulation experiments are compared with advanced algorithms to prove the effectiveness of the improved algorithm.(4)According to the improved algorithm presented in this paper,it is combined with the real-time SLAM system and tested in the real environment.Through comparison with the original algorithm,it proves that the improved method has improved both in time consuming and the result of map construction.